Hi Miles, The gcc versions on two exec nodes are different. bunix-server has 4.4.5 and the other 4.6.1. Both are 32 bits Ubuntu Desktop, one 10.10 and the other 11.10. The headnode is installed with 64 bits Ubuntu Server 11.10. The headnode has also 4.6.1 gcc.
I know Moses is written for 32 bits systems and can be compiled on 64 bits systems. I don't know whether this will change the code to 64 bits. I just rechecked the out.job12017-aa/b files. They are a bit different. I put them all together here for comparison. out.job12017-aa: Linux bunix-server 2.6.35-30-generic #60-Ubuntu SMP Mon Sep 19 20:45:08 UTC 2011 i686 GNU/Linux ulimit: Command not found. /home/guchun/Work/mosesdecoder/moses-cmd/src/moses: Exec format error. Wrong Architecture. Newline in variable name. bunix-server is Ubuntu 10.10. out.job12017-ab: Warning: no access to tty (Bad file descriptor). Thus no job control in this shell. Linux guchun-VirtualBox 3.0.0-12-generic #20-Ubuntu SMP Fri Oct 7 14:50:42 UTC 2011 i686 athlon i386 GNU/Linux ulimit: Command not found. /home/guchun/Work/mosesdecoder/moses-cmd/src/moses: Exec format error. Binary file not executable. exit status 1 mv: cannot stat `/home/guchun/Work/tasks/ro-en/tuning-sge/tmp12017/run1.best100.out.split12017-ab': No such file or directory exit status 1 exit status 0 guchun-VirtualBox is Ubuntu 11.10, installed in a VirtualBox on bunix-server. The error messages are different. Is this caused by the different version of Ubuntu? Or something more profound? Cheers, Guchun On 16 November 2011 12:21, Miles Osborne <[email protected]> wrote: > check the gcc version on the slaves. it looks like eg you are running > 64 code on a 32 bit machine > > Miles > > On 16 November 2011 12:15, Guchun Zhang <[email protected]> wrote: > > Hi Barry, > > In out.job12017-aa, > > Linux bunix-server 2.6.35-30-generic #60-Ubuntu SMP Mon Sep 19 20:45:08 > UTC > > 2011 i686 GNU/Linux > > ulimit: Command not found. > > /home/guchun/Work/mosesdecoder/moses-cmd/src/moses: Exec format error. > Wrong > > Architecture. > > Newline in variable name. > > bunix-server is the hostname of the execution node. Complaints are > similar > > in out.job12017-ab (run on another node), too. > > Cheers, > > Guchun > > On 16 November 2011 09:21, Barry Haddow <[email protected]> wrote: > >> > >> Hi Guchun > >> > >> The mert.out file doesn't help that much. Is there any more information > in > >> the > >> err and out files? > >> eg > >> /home/guchun/Work/tasks/ro-en/tuning-sge/out.job12017-aa > >> /home/guchun/Work/tasks/ro-en/tuning-sge/err.job12017-aa > >> > >> cheers - Barry > >> > >> On Tuesday 15 Nov 2011 22:01:41 Guchun Zhang wrote: > >> > Hi there, > >> > > >> > I am trying to tune on a SGE cluster. I ran the following command on > the > >> > head node, > >> > > >> > > >> > > /home/guchun/Work/moses-scripts/scripts-20111111-1703/training/mert-moses.p > >> > l \ > >> > /home/guchun/Work/tasks/ro-en/corpus/euparl.lc.ro \ > >> > /home/guchun/Work/tasks/ro-en/corpus/euparl.lc.en \ > >> > /home/guchun/Work/mosesdecoder/moses-cmd/src/moses \ > >> > /home/guchun/Work/tasks/ro-en/trained/model/moses.ini \ > >> > --mertdir /home/guchun/Work/mosesdecoder/mert/ \ > >> > --rootdir /home/guchun/Work/moses-scripts/scripts-20111111-1703/ \ > >> > --working-dir /home/guchun/Work/tasks/ro-en/tuning-sge/ \ > >> > --jobs 2 --decoder-flag "-v 0" >& > >> > /home/guchun/Work/tasks/ro-en/tuning-sge/mert.out & > >> > > >> > I got the following error, > >> > > >> > check_exit_status > >> > check_exit_status of job -aa > >> > check_exit_status of job -ab > >> > *wc: euparl.lc.ro.split12017-aa.trans: No such file or directory* > >> > *Split (-aa) were not entirely translated* > >> > outputN= inputN=11966 > >> > outputfile=euparl.lc.ro.split12017-aa.trans > >> > inputfile=euparl.lc.ro.split12017-aa > >> > *Split (-ab) were not entirely translated* > >> > outputN=0 inputN=11966 > >> > outputfile=euparl.lc.ro.split12017-ab.trans > >> > inputfile=euparl.lc.ro.split12017-ab > >> > *everything crashed, not trying to resubmit jobs* > >> > *Got interrupt or something failed.* > >> > kill_all_and_quit > >> > qdel 56 > >> > Executing: qdel 56 > >> > Exit code: 1 > >> > qdel 57 > >> > Executing: qdel 57 > >> > Exit code: 1 > >> > Translation was not performed correctly > >> > or some of the submitted jobs died. > >> > qdel function was called for all submitted jobs > >> > Exit code: 1 > >> > The decoder died. CONFIG WAS -w -0.322581 -lm 0.161290 -d 0.193548 -tm > >> > 0.064516 0.064516 0.064516 0.064516 0.064516 > >> > > >> > Any clue what may cause the problem? I have also attached the output > >> > file > >> > (mert.out) for full inspection. > >> > > >> > Everything runs fine in serial execution (without --job 2). > >> > > >> > I wonder if this can attribute to my SGE configuration. So if > possible, > >> > could you please also give some advice on the parameter configuration > of > >> > SGE? > >> > > >> > Many thanks in advance, > >> > > >> > Guchun > >> > > >> > > > > > > > > -- > > > > Guchun Zhang > > > > Localization Engineer > > Alpha CRC Ltd | Cambridge, UK > > Direct: +44 1223 431035 > > [email protected] > > > > _______________________________________________ > > Moses-support mailing list > > [email protected] > > http://mailman.mit.edu/mailman/listinfo/moses-support > > > > > > > > -- > The University of Edinburgh is a charitable body, registered in > Scotland, with registration number SC005336. > > -- *Guchun Zhang* Localization Engineer Alpha CRC Ltd | Cambridge, UK Direct: +44 1223 431035 [email protected] <[email protected]>
_______________________________________________ Moses-support mailing list [email protected] http://mailman.mit.edu/mailman/listinfo/moses-support
